Introduction

Beginning in 1993, researchers from the Andrews Experimental Forest Long-Term Ecological Research (LTER) have been monitoring populations of Pacific Giant Salamanders, Dicamptodon tenebrosus, specifically in Mack Creek (Gregory 2016). The Experimental Forest is part of the Willamette National Forest in western Oregon and managed by the U.S. Forest Service and Oregon State University (Figure 2). Researchers are interested in how disturbances in forests effect biota. The data was collected from two sections of Mack Creek - a forest clear cut in 1963 and a 500 year old growth coniferous forest. Here, we will use the data collected by the Andrews Experimental Forest Team to compare the differences in size and weight distributions of salamanders from clear cut and old growth sections.

Figure 1: A Pacific Giant Salamander (Dicamptodon tenebrosus) laying on a down log. Photo credit: Gary Nafis, www.californiaherps.com.

Figure 2: Map of HJ Andrews Experimental Forest LTER in Oregon. Andrews Experimental Forest is managed by the U.S. Forest Service and Oregon State University. Map courtesty of: HJ Andrews Experimental Forest.

Data and Methods

Data for salamander counts and weights were annually collected since 1993 in two 150 meter reaches within Mack Creek. The lower reach is in a clear cut forest, while the upper reach is in an old growth forest. The reaches were divided into three 50 meter sections blocked off by nets, then each 50 m section was sampled with two-pass electrofishing (Gregory 2016). All vertebrates captured were measured and weighed. Channel measurements, such as channel type, were aso recorded. For more information on collection methods and spatial information, please refer to the metadata.

Trends in salamander counts from 1993-2017 were visually compared in clear cut and old growth reaches. Salamander counts were then compared using a Pearson’s chi-square test to determine if salamander counts differed statistically by channel type (cascade, pool, side channel; \(\alpha\) = 0.05 throughout). 2017 salamander weights were compared in clear cut and old growth forest sections and in three different channel types using a two sample t test and an ANOVA, respectively, to determine if mean weights differed sigifnicantly by reach section and or channel type. To measure the maginitude of the difference between the means, we calculated the effect size using Cohen’s d. All analyses and figures were prepared using R software version 3.6.1. Collaboration between authors was done using GitHub.

Results

Results A: Annual salamander counts in old growth and clear cut sections of Mack Creek

Salamander counts in old growth and clear cut sections were compared between 1993 and 2017. Overall, salamander counts in both sections followed similar patterns; when counts in old-growth sections increased, counts in clear cut sections also increased. Similarly, when counts of salamanders in old-growth sections decreased, salamander counts in clear cut sections also decreased. Likely, environmental factors affect annual variability in salamander counts.

Generally, salamander observations were higher in old growth forest. However, in 1994 and 2015-2017, salamander counts were higher in clear cut sections. Between 1993 and 2017, salamander counts in the clear cut and od growth forests increased by 129% and 154%, respectively. Throughout the study period, salamander counts in the clear cut section reached a high of 368 in 2017 and a low of 137 in 2014. Comparatively, salamander counts in the old growth section reached a high of 380 in 2002 and a low of 129 in 1993. In general, salamader counts have been increasing since the study began in 1993.

Figure 3: Annual salamander counts in clear cut (orange) and old growth (green) sections of Mack Creek. Old growth sections are responsible for higher counts of salamanders for 80% of the time between 1993 and 2017. The populations in both sections display similar patterns of highs and lows. Data: Stanley Gregory, Andrews Forest LTER Sites.

Results B: Table of 2017 salamander counts by channel classification

Table 1: Salamander counts in different channel types (cascade, pool, and side channel) in clear cut and old-growth forests. Data: Stanley Gregory, Andrews Forest LTER Sites.
Channel Type Clear Cut Forest Old-Growth Forest
Cascade 247 (55.1%) 201 (44.9%)
Pool 31 (40.8%) 45 (59.2%)
Side Channel 90 (54.9%) 74 (45.1%)

Salamander counts in clear cut sections were higher in cascades (ncascade = 247 (55.1%)) and side channels (nsidechannel = 90 (54.9%)). In old-growth sections, pools had higher counts of salamanders (npool = 45 (59.2%)).

Results C: Is there a significant difference in salamander counts in channel locations in old growth and clear cut forests?

We performed a Pearson’s chi-squared test to examine whether or not there was a significant difference in where in the channel Pacific giant salamanders are located between the two sections of forest (old-growth and clear cut). The results from our test found that there was no siginificant difference in what channel Pacific giant salamander are located (\(\chi\)2(2) = 5.54, p > 0.001). This can be seen in Table 1 - the counts in clear cut and old-growth channel types were not different enough for the chi-squared test to find a significant difference. For example, Salamander counts were actually higher in clear cut sections in the forest for side channels (n~side channel~ = 90) and cascade (ncascade = 247) channel types. Pools were the only channel type to gave higher salamander counts in old-growth forests (npool = 45).

From the results of Table 1 and the chi-squared test, it seems evident that salamanders can adapt to the clear-cutting of forests and can possibly survive better than in intact old-growth forests.

Results D: Is there a significant difference in mean weights for salamanders observed in clear cut vs old growth forests in 2017?

A two-sample t-test was used to determine if mean salamander weights in 2017 were significantly different in old-growth and clear cut forests. 2017 salamander weights in clear cut forests (7.78g) did not differ significantly from 2017 salamander weights in old-growth forests (6.58g) (t(692.79) = 1.67, p = 0.1). A negligable effect size (0.1) confirms that there is no difference in mean 2017 weights by section along Mack Creek.

Results E: Is there a significant difference between mean salamander weights in the 3 channel types of Mack Creek in 2017?

Salamanders found in side channels of Mack Creek had the lowest mean weight (mean2017 = 5.68 grams) compared to salamanders found in pools (mean2017 = 9.30 grams) and cascades (mean2017 = 7.52 grams) (Figure 4.).

Figure 4: 2017 weights of salamanders in cascade, pool, and side channels of Mack Creek. Mean weights for each channel type are indicated by the black center dot. Standard deviation of weights are indicated by the upper and lower bars. Data: Stanley Gregory, Andrews Forest LTER Sites.

To determine whether or not there is a difference in 2017 mean weight for salamanders found in different channel types (pool, cascade, and side channel), we performed a one-way ANOVA test. The results indicate that there is a significant difference between at least two of the mean weights for salamanders in different channels. A Tukey’s HSD post-hoc test was performed to determine which channels had significant differences between 2017 mean weights F(2, 684) = 4.22, p = 0.015). The results show that there is a significant difference in one pairwise comparison: mean weights between salamanders in side channels (5.68g \(\pm\) 0.65g)(mean \(\pm\) SE) and pools (9.30g \(\pm\) 1.56g)(mean \(\pm\) SE). There was no significant difference in mean weights of salamanders between pools (9.30g \(\pm\) 1.56g)(mean \(\pm\) SE) and cascades (7.52g \(\pm\) 0.43g)(mean \(\pm\) SE), as well as no significant difference between side channel (5.68g \(\pm\) 0.65g)(mean \(\pm\) SE) and cascades (7.52g \(\pm\) 0.43g)(mean \(\pm\) SE).

The actual differences in mean weights between channel types was 1.78 grams between pools and cascades and 1.84 grams between side channels and cascades. These differences were significantly lower than the difference between pools and side channles, which was 3.62 grams (about 50% higher than the others).

Our above tests are parametric tests that include assumptions of normal distribution of the salamander populations, which was done so by using mean weights of salamanders. However, an unpaired non-parametric test may also be appropriate. Non-parametric tests can be used for non-normal distributions, as we saw our salamander data exhibited above through qq tests and histograms. In that case, we would perform a Mann-Whitney U test to rank the weights of the salamanders instead of comparing the means.

Summary

References

HJ Andrews Experimental Forest Long Term-Ecological Research. https://andrewsforest.oregonstate.edu/sites/default/files/lter/data/map/2011base.jpg

Gregory S. V. 2016. Aquatic Vertebrate Population Study in Mack Creek, Andrews Experimental Forest, 1987 to present. Environmental Data Initiative. https://doi.org/10.6073/pasta/5de64af9c11579266ef20da2ff32f702. Dataset accessed 11/25/2019.

Nafis, Gary. “Great Pacific Salamander.” CaliforniaHerps. http://www.californiaherps.com/salamanders/pages/d.tenebrosus.html